The scrap rate is rising, but no one knows when the process has gone off track. The problem isn’t detected until the final inspection—by which time the defective parts have long since been produced. SPC is designed to break this very pattern: to detect deviations as they occur, not hours later. The difference between these two operating modes determines whether a malfunction results in the loss of a few parts or an entire batch.
Statistical process control is an old tool that is often misused. Control charts are posted but not read. Action limits are confused with tolerance limits, causing the entire early-warning function to be lost. And data collection remains so slow that a trend isn’t recognized until it has long since had an impact. When used correctly, however, SPC is the most effective means of preventing systematic scrap.
This article shows how SPC really works in manufacturing: from the control chart and its warning signals, through the crucial difference between action limits and tolerance limits, to the question of how SPC and Cpk are related. By the end, you’ll understand how simple process control evolves into predictive process control—and what data foundation is required for that.
KEY POINTS AT A GLANCE
|
IN SHORTSPC, or statistical process control, monitors manufacturing processes using control charts and distinguishes between random and systematic deviations. At its core is the control chart with its action limits derived from the process, which must not be confused with the tolerance limits specified in the requirements. SPC provides the data basis for Cpk and only becomes a true control tool through automatic data collection. |
|
CONTENTS OF THIS ARTICLE |
What is SPC in manufacturing?
SPC stands for Statistical Process Control. It is a method that continuously monitors processes and distinguishes between random and systematic deviations. Random variation is normal and is tolerated. Systematic deviation indicates an underlying cause and requires corrective action.
The goal is prevention rather than correction. Instead of inspecting finished parts and sorting out rejects, SPC monitors the process itself and alerts you when it changes. This allows errors to be corrected before they occur on a large scale.
| Type of Variation | Cause | Response |
|---|---|---|
| Random (Common Cause) | Natural process variation | No action required; tolerate |
| Systematic (Special Cause) | Tool wear, material change, error | Identify and eliminate the cause |
At the heart of SPC is a simple but far-reaching distinction: Not every variation is a problem. Every process inherently exhibits variation, and this random noise should not trigger corrective action, because every unnecessary intervention actually increases the variation. SPC provides an objective criterion for determining when corrective action is warranted and when it is not.
The control chart as the centerpiece of SPC
The control chart is the central tool of SPC. It plots measured values over time and uses control limits to define the range within which a controlled process is allowed to vary. Points within the limits are unremarkable; points outside the limits indicate a systematic deviation.
In addition to out-of-control points, patterns also provide clues: seven consecutive points on one side of the centerline, a sustained trend, or a cycle indicate a systematic cause, even if no point crosses the limit.
| Patterns | Meaning | Example cause |
|---|---|---|
| Point Outside the Limit | Systematic deviation | Tool breakage |
| Seven points on one side | Mean value shift | New material batch |
| Sustained trend | Gradual drift | Tool wear |
| Cyclical pattern | Periodic influence | Temperature fluctuation |
The control limits are calculated from the process itself, typically as the mean plus or minus three standard deviations of the subgroup distribution. They therefore describe how the process behaves, not how it should behave. This is precisely where their early-warning capability lies: they are triggered as soon as the process deviates from its usual behavior, long before a part violates the tolerance.
The selection of the correct control chart depends on the characteristic. For variable measurements, mean and range charts are used; for attributive characteristics, frequency charts are used. Choosing the wrong type of chart leads to incorrect control limits and thus to false alarms or overlooked deviations. Choosing the right chart is therefore not a mere formality, but a technical decision.
Distinguishing Between Intervention Limits and Tolerance Limits
The most common mistake in SPC is confusing action limits with tolerance limits. Both look similar on the chart, but they mean opposite things.
| Characteristic | Action Limit | Tolerance Limit |
|---|---|---|
| Origin | Calculated from the process | Specified in the specification |
| Purpose | Monitor the process | Evaluate product |
| Change | If the process changes | When the drawing changes |
| On the control chart | Yes | No, it doesn't belong there |
Control limits provide information about the process, while tolerance limits provide information about the product. A process can exceed its control limits and still remain within tolerance, and vice versa. Plotting tolerance limits on the control chart dilutes the very early-warning function that defines SPC.
Tolerance limits on a control chart are a contradiction in terms. The control chart monitors the process, not the product. Mixing the two defeats the purpose of early warning.
Amadeus Lederle, Chief Technology Executive, CSP Intelligence GmbH
The consequences of this confusion are serious. Anyone who uses tolerance limits as action limits only reacts once the process is already producing scrap. The entire preventive function of SPC is lost, and the control chart becomes a means of documenting errors after the fact rather than an early warning system. Action limits are always narrower than tolerance limits; otherwise, the process would not be capable.
Another reason against using tolerance limits on the control chart is statistical in nature. Control limits refer to the distribution of subgroup means, while tolerance limits refer to the distribution of individual values. Both have different widths and reference values. Mixing them on the same chart compares quantities that are not comparable and leads to incorrect conclusions.
From Manual to Automatic Data Collection
SPC stands or falls on data collection. If values are read manually and entered onto a paper chart, delays and transcription errors occur. By the time a manually maintained chart shows a critical trend, many parts have often already been manufactured.
| Aspect | Manual | Automatic |
|---|---|---|
| Data Collection | Read and record | Directly from the measuring device |
| Response time | Minutes to hours | Seconds |
| Source of error | Transmission | Virtually none |
| Suitability | Initial analysis | Ongoing program |
Manual SPC is sufficient for an initial analysis or a single investigation. For a long-term program involving many characteristics and lines, it is too slow and too prone to errors. At this point, there is no way around automated data collection.
The distinction between manual and automated data collection determines the system’s usefulness. A hand-filled chart that is evaluated once per shift does not detect a trend until hours later. Automated data collection evaluates every value immediately and reports deviations within seconds. In that amount of time, the difference can be between a few rejected parts and an entire batch.
How SPC and Cpk Are Related
SPC and Cpk are two sides of the same coin. SPC monitors whether a process is stable. Cpk assesses whether a stable process is also capable. Without SPC, there can be no valid Cpk, because Cpk requires a controlled process.
The control chart provides the short-term variation from which Cp and Cpk are calculated. Only when stability has been demonstrated through SPC does the Cpk become meaningful.
PRACTICAL NOTEA Cpk without accompanying SPC is a snapshot with no predictive value. Continue to maintain the control chart on an ongoing basis, even after capability has been demonstrated. This is the only way to detect when a process that was once capable begins to drift. |
The order is mandatory here as well: stability first, then capability. SPC verifies stability; the Cpk requires it. A Cpk derived from a process without an accompanying control chart is not reliable, because no one knows whether the process was under control at the time of measurement. SPC and Cpk are therefore not parallel processes, but sequential ones.
In practice, SPC and capability assessment ideally converge within the same system. The control chart continuously provides the variation data from which the capability indices are calculated without the need for additional data entry. This results in a seamless workflow from measurement through stability testing to the capability assessment, without any media breaks or duplicate data entry.
From monitoring to true process control
Traditional SPC is steering: It signals a deviation, and a person responds. The next level is control: The system detects a trend and corrects the process before the limit is reached. The difference lies in the response time and the degree of automation.
| Level | Detection | Reaction | Benefits |
|---|---|---|---|
| Process Control | Control chart shows deviation | Human intervenes | Less scrap |
| Process control | System detects trends early | System or operator corrects in advance | Virtually no scrap |
The transition from manual control to automated control is also the transition from experience to a systematic approach. In manual control, the operator makes decisions based on the chart; in automated control, the system detects trends and either suggests or implements corrections. Human experience remains valuable but is supplemented by objective, seamless monitoring.
The transition to control requires a seamless, rapid data foundation. As long as values are recorded manually, only manual guidance is possible because the response time is too long. Only automatic data collection creates the foundation on which a system can detect trends and make proactive corrections. Control is thus less a matter of will than of data infrastructure.
SPC as a Continuous Program
An SPC program covering many characteristics, machines, and shifts cannot be managed manually over the long term. The CSP Manufacturing OS captures measurement data directly from the measuring equipment, automatically maintains control charts, and calculates capability indices in real time. The IPM module detects trends and out-of-control points, triggers alerts, and archives each control chart in an audit-proof manner. This transforms simple process control into predictive process control.
The scalability advantage of a system-supported SPC becomes apparent only when applied at scale. A single control chart can be managed manually. However, manually managing hundreds of charts across numerous machines and shifts—each with accurate trend detection and audit-traceable storage—is simply not feasible. This is precisely where a good tool becomes a sustainable program.
Frequently Asked Questions
What Is SPC in Manufacturing?
SPC stands for Statistical Process Control. It monitors manufacturing processes using control charts and distinguishes between random and systematic deviations to detect errors before scrap is produced.
What is the difference between an action limit and a tolerance limit?
Control limits are calculated from the process and are used to monitor the process. Tolerance limits are derived from the specification and are used to evaluate the product. Tolerance limits do not belong on the control chart.
How are SPC and Cpk related?
SPC demonstrates that a process is stable and provides data on variation. The Cpk assesses whether the stable process is also capable. Without the stability demonstrated by SPC, the Cpk has no predictive value.
What warning signals does a control chart show?
In addition to points outside the control limits, certain patterns serve as warnings: seven consecutive points on one side of the centerline, a sustained trend, or a cyclical pattern indicate a systematic cause.
Is manual SPC sufficient?
Manual SPC is sufficient for an initial analysis or a single investigation. For a continuous program involving many characteristics, it is too slow and too prone to errors; in such cases, automatic data collection is necessary.
What is the difference between process control and process regulation?
Process monitoring signals a deviation, to which a person responds. Process control detects a trend early and corrects the process before a limit is reached. The difference lies in response time and automation.
Which standard governs SPC?
In the automotive industry, SPC is governed by the AIAG-VDA SPC Guide and IATF 16949. The ISO 22514 series of standards also applies to statistical methods.